Quantitative Multiscale Analysis using Different Wavelets in 1D Voice Signal and 2D Image

TL;DR

This paper conducts a multiscale analysis of 1D voice signals and 2D images using various wavelets and transforms, evaluating signal quality and compression performance at multiple scales.

Contribution

It provides a comprehensive quantitative comparison of different wavelets and transforms for multiscale analysis and image compression performance evaluation.

Findings

Wavelet-based multiscale analysis improves signal quality metrics.

Image compression rates exceed 92% with specific wavelets.

Different wavelets show varying effectiveness in analysis and compression.

Abstract

Mutiscale analysis represents multiresolution scrutiny of a signal to improve its signal quality. Multiresolution analysis of 1D voice signal and 2D image is conducted using DCT, FFT and different wavelets such as Haar, Deubachies, Morlet, Cauchy, Shannon, Biorthogonal, Symmlet and Coiflet deploying the cascaded filter banks based decomposition and reconstruction. The outstanding quantitative analysis of the specified wavelets is done to investigate the signal quality, mean square error, entropy and peak-to-peak SNR at multiscale stage-4 for both 1D voice signal and 2D image. In addition, the 2D image compression performance is significantly found 93.00% in DB-4, 93.68% in bior-4.4, 93.18% in Sym-4 and 92.20% in Coif-2 during the multiscale analysis.